Published ahead of print on May 18, 2005, doi:10.1164/rccm.200411-1498OC
© 2005 American Thoracic Society doi: 10.1164/rccm.200411-1498OC
Exhaled Nitric OxideA Predictor of Steroid ResponseOtago Respiratory Research Unit, Dunedin School of Medicine, University of Otago Medical School, Dunedin, New Zealand Correspondence and requests for reprints should be addressed to Professor D. Robin Taylor, M.D., F.R.C.P., Dunedin School of Medicine, University of Otago Medical School, P.O. Box 913, Dunedin, New Zealand. E-mail: robin.taylor{at}stonebow.otago.ac.nz
Rationale: The initial management of patients who present with persistent respiratory symptoms includes recognizing those with the potential to benefit from inhaled steroid therapy. To date, this has required undertaking a "trial of steroid" to identify responders. There is increasing evidence that steroid response is more likely in patients with eosinophilic airway inflammation, and this can be assessed indirectly using exhaled nitric oxide (FENO) measurements. Objectives: We aimed to assess the predictive accuracy of FENO to identify steroid response in 52 patients presenting with undiagnosed respiratory symptoms in a single-blind, fixed-sequence, placebo-controlled trial of inhaled fluticasone for 4 weeks. Methods: Comparisons of predictive accuracy were made between FENO and other conventional predictors: peak flows, spirometry, bronchodilator response, and airway hyperresponsiveness measured at baseline. "Steroid response" was defined as change in symptoms, peak flows, spirometry, or airway hyperresponsiveness to adenosine based on established guidelines and recommendations. Results: Steroid response was significantly greater in the highest FENO tertile (> 47 ppb) for each endpoint. This outcome was independent of the diagnostic label. The predictive values for FENO were significantly greater than for almost all other baseline predictors, with an optimum cut point of 47 ppb. Conclusions: FENO measurements greater than 47 ppb provide a means of predicting steroid response in patients with undiagnosed respiratory symptoms. Assessing airway inflammation is of more practical value than diagnostic labeling when considering the potential usefulness of inhaled antiinflammatory therapy.
Key Words: asthma • exhaled nitric oxide • inhaled corticosteroid • symptoms • treatment response Managing patients with chronic respiratory symptoms due to airway pathology is commonplace, particularly in the primary care setting. Understandably, clinicians aim to establish a firm diagnosis in the first instance, and then proceed to offer treatment. This will often include inhaled antiinflammatory therapy, particularly if the diagnosis is believed to be asthma. However, even in the absence of a confirmed diagnosis, inhaled corticosteroids (ICSs) are often prescribed empirically. This approach may be problematic. First, respiratory symptoms are nonspecific, with considerable overlap between conditions such as asthma, wheezy bronchitis, eosinophilic bronchitis, postviral airway hyperresponsiveness, chronic obstructive pulmonary disease, vocal cord dysfunction, and primary hyperventilation syndrome. Other related problems may contribute to confusion regarding the true etiology of the patient's symptoms (e.g., gastroesophageal reflux, chronic rhinitis with postnasal drip) (1). Second, conventional lung function tests (e.g., spirometry, peak flows) are poorly sensitive at least for the diagnosis of asthma (2), and cannot be relied on to provide supportive evidence, especially in mild cases. Even if the diagnosis of asthma is firm, there is significant heterogeneity in the response to ICS treatment (3). Thus, patients may be committed to steroid therapy inappropriately, and this is both costly and potentially harmful, particularly at higher doses (4, 5). An alternative approach would be to identify "steroid responsiveness" in relation to underlying airway inflammation as part of the first-line assessment of patients with undiagnosed respiratory symptoms, in addition to and independently of attaching a diagnostic label. This has already been suggested for chronic obstructive pulmonary disease (6), and given that there is significant overlap between chronic obstructive pulmonary disease and asthma (7), attaching a diagnostic label per se is not necessarily helpful in predicting steroid responsiveness (8). There is increasing evidence that steroid response is more likely in patients with eosinophilic airway inflammation (9), and this can be assessed, albeit indirectly, using exhaled nitric oxide (FENO) as a surrogate measurement (10–12). With ICS treatment, FENO levels are reduced in a dose-dependent manner (13, 14), and conversely, levels increase when ICS treatment is withdrawn (15). Taken together, these data suggest the possibility that FENO measurements might be used to predict whether patients with nonspecific respiratory symptoms will respond to a trial of inhaled steroids. In this single-blind, fixed-sequence, placebo-controlled study, our aim was to evaluate the accuracy of baseline FENO measurements to predict a positive response to treatment with inhaled fluticasone.
Subjects We recruited consecutive patients referred by their family practitioner to the Dunedin Hospital pulmonary function laboratory for investigation of persistent, previously undiagnosed respiratory symptoms. The laboratory offers spirometry and bronchial challenge testing as a routine service for community practitioners. Patients were 12 to 75 years old and had respiratory symptoms for a minimum of 6 weeks. Exclusion criteria were as follows: use of ICSs or oral corticosteroids in the previous 4 weeks, respiratory tract infection in the previous 6 weeks, other established respiratory diagnosis, or significant comorbidity. Smokers were not excluded. The study was approved by the Otago Ethics Committee, and all subjects gave written, informed consent.
Study Design
Study Procedures Respiratory symptoms, bronchodilator use, and peak flows were recorded twice daily in a diary during the run-in and each treatment period. Diurnal peak flow variation was calculated as amplitude % mean over 7 days: a value of 20% or greater was considered clinically significant (20). Using the diary data, a composite symptom score (0 to 10) was calculated for each day (see online supplement). Mean values for the composite symptom scores and morning peak flows were obtained for three 7-day intervals: before commencing placebo treatment (run-in), last 7 days of placebo treatment period, and last 7 days of fluticasone treatment period. FENO was measured using a chemiluminescence analyzer (NiOX; Aerocrine, Stockholm, Sweden) before any forced expiratory maneuvers according to current guidelines at an exhaled flow rate of 50 ml/second (21). All readings were recorded by two staff members who were blinded to the patient's identity and treatment period. Bronchodilator reversibility was calculated as the percentage of change in FEV1 from baseline, 15 minutes after inhaling 400 µg of albuterol. PD20 methacholine and provocative concentration (PC) of adenosine monophosphate (AMP) resulting in a 20% reduction in FEV1 (PC20) were measured using standard protocols (22, 23).
Study Measurements and Statistical Analyses
One hundred one consecutive patients were referred to the pulmonary function laboratory during the recruitment interval; of these patients, 60 were enrolled into the study. Thirteen patients could not be contacted further; 13 declined to participate; 15 did not meet the inclusion criteria (7 were taking ICSs, 4 did not have chronic symptoms, 2 had significant comorbidities, 1 had sarcoidosis, and 1 had lower respiratory tract infection). During the study, six patients failed to complete the placebo treatment period (one withdrew consent, one had respiratory tract infection, one had acute rhinosinusitis, and three were lost to follow-up). Two patients withdrew during the fluticasone treatment period (both withdrew consent). Demographic details for the 52 patients who completed the study are shown in Table 3, and have been further stratified for baseline FENO by tertiles.
A diagnosis of asthma was made in 27 of the 52 patients. The criteria used to confirm the diagnosis of asthma in individual patients are reported in Table E1 in the online supplement. The proportion of patients with asthma was greatest in the highest FENO tertile (15/17, 88%) compared with the middle (7/18, 39%) and lowest tertiles (5/17, 29%; p < 0.001). In the remaining 25 patients, the final diagnoses were as follows: postnasal drip (13), bronchitis (4), gastroesophageal reflux disease (4), primary hyperventilation syndrome (2), bronchiectasis (1), and left ventricular dysfunction (1). Subjects with baseline FENO levels in the highest tertile (> 47 ppb) had significantly lower FEV1 % predicted and FEV1/FVC ratio and significantly greater improvement in FEV1 with bronchodilator compared with the other two tertiles. Subjects with FENO in the highest tertile also had a significantly greater response to inhaled fluticasone for all four categories of "steroid response" (increase in FEV1, increase in mean morning peak flows, improved respiratory symptoms, and reduction in airway hyperresponsiveness to AMP) as shown in Table 3 and Figure 1.
ROC curves for each of the baseline measurements used as predictors of steroid response are shown in Figure 2. The AUCs for baseline FENO, FEV1 % predicted, FEV1 bronchodilator response, PD20 methacholine, and peak flow variation used as predictors were as follows: 0.76, 0.47*, 0.66, 0.63, and 0.63 for increase in FEV1; 0.81, 0.68, 0.57*, 0.65*, and 0.58*for increase in morning peak flow; 0.64, 0.45*, 0.46*, 0.45*, and 0.46 for decrease in symptom score; and 0.91, 0.68*, 0.71*, 0.84, and 0.72* for improvement in PC20 AMP. Values with an asterisk indicate those predictors for which comparisons with FENO revealed significant differences (see Tables E2–E5 for z-scores and p values). Sensitivities, specificities, and positive and negative predictive values for each of the predictors in relation to each of the four "steroid response" outcomes are shown in Table 4. In general, baseline FENO provided greater sensitivities and negative predictive values than for each of the other predictors. Overall, the optimum FENO cut point as a predictor for each of the four study endpoints was 47 ppb (see Tables E6–E9). This cut point was obtained independently of the cut point for the highest FENO tertile.
The results of the present study confirm that using a wide range of possible outcomes to define "steroid response," FENO measurements provide superior accuracy compared with the majority of other predictors for identifying "responders" among patients with previously undiagnosed persistent respiratory symptoms. First, the change in each of the major study outcomes after 4 weeks of inhaled fluticasone was significantly greater in the highest FENO tertile group. Second, the AUCs for FENO were consistently greater than for any other predictor of steroid response (FEV1 % predicted, bronchodilator reversibility, airway hyperresponsiveness to methacholine, and peak flow variation). For FENO, the negative predictive values for steroid response ranged from 77 to 94% depending on the endpoint chosen (Table 4), indicating that, in the absence of a high FENO (> 47 ppb), steroid responsiveness is less likely to be present. We have previously demonstrated that FENO measurements may be used helpfully to diagnose asthma (2). Furthermore, the combination of FENO and spirometric testing provides even higher levels of diagnostic accuracy (25). However, in practice, it is arguably just as important to identify patients who may or may not benefit from inhaled antiinflammatory therapy as it is to attach a diagnostic label. This has been confirmed in the present study. As we have demonstrated, steroid response occurred in relation to baseline FENO measurements independently of the final diagnosis. This highlights the limitations of using diagnostic labeling to predict treatment response. Rather, it emphasizes the need to assess airway inflammation, even indirectly, to direct treatment (6, 9). Measuring FENO is potentially very useful in this regard given that it correlates well with eosinophilic airway inflammation (11, 12). Previously, Pavord and colleagues (9) have confirmed that, in a group of 23 patients with asthma, the response to a short-term trial of inhaled budesonide was greatest in patients with significant sputum eosinophilia. A similar outcome was reported by Meijer and coworkers (26) who studied 120 patients with asthma whose inhaled steroid therapy was withdrawn. High levels of sputum eosinophils were predictive of steroid response. This underlines that, even within a group of patients with diagnosed asthma, steroid response is related to particular characteristics of airway inflammation. Little and colleagues (27) have also confirmed that in patients with asthma, most of whom were taking inhaled steroids, both FENO and sputum eosinophil counts had high predictive values for any additional available improvement in airway caliber with a trial of oral prednisolone. However, close comparisons with the present study are difficult, given the differences in selection criteria for our study population as well as in the cut points for both FENO and steroid response measurements. In the study by Meijer and coworkers (26), FENO levels were not found to provide predictive accuracy, but this may have been influenced by differences in the methods used for FENO measurement. The strength of the present study is that unselected, steroid-naive patients both with and without asthma were enrolled. This permitted the assessment of airway inflammation de novo (albeit indirectly) to be confirmed as a means of predicting steroid response. Our study design was influenced by a number of factors. First, it has been shown that with a trial of inhaled or oral steroid, a significant carryover effect is likely to occur (8, 28). To avoid this would have required a washout period of at least 1 month. Because our patients were presenting with undiagnosed, untreated symptoms, it was not considered ethical to extend the study (to 3 months) to conduct a random-order crossover study. Rather, a fixed-sequence, single-blind design was used, with placebo given during the first treatment period. In practice, a "trial of steroid" does not usually include placebo comparisons. In our study, the within-treatment placebo response was subtracted from the within-treatment response to fluticasone. This enabled issues of regression to the mean as well the true placebo responses to be taken into account. We chose to report the relationship between steroid response outcomes and baseline FENO levels using tertiles. This was to facilitate comparisons with a previous study (6). Analysis of our data using FENO either as a continuous variable or divided into smaller groups (e.g., quintiles) did not alter the statistical significance of the association between FENO levels and study outcomes. Clearly, our results were dependent on the criteria used to define "steroid response." These were derived from current international guidelines (17–19). Using these criteria, the specificities for a significant steroid response using FENO as predictor ranged from 71 to 91%, implying that up to 30% of patients with a high FENO failed to demonstrate steroid responsiveness (Table 4). This may be explained in part by the fact that many patients had normal lung function, and there was therefore little room for significant improvement with fluticasone. The cut points used to define "response" may therefore have been inappropriate in these patients. Arguably, such individuals might still benefit from inhaled steroid treatment using criteria for response that were not used in our study. For example, we did not include longer term improvement in symptoms or quality of life as part of our study protocol. In conclusion, our data provide further evidence of the usefulness of FENO measurements in clinical practice. The likelihood of steroid responsiveness, or perhaps unresponsiveness, may be helpfully predicted using single FENO measurements in patients with previously undiagnosed respiratory symptoms. The cut point for optimum predictive accuracy was 47 ppb, which is notably higher than the upper limit of the so-called normal range (35 ppb) (29). This finding does not preclude the possibility that steroid responsiveness may occur at lower FENO levels, but clearly this is less likely. Using FENO measurements provides an alternative to securing and applying a diagnostic label, particularly that of asthma, as a guide to treatment. Of course, there is still the need to arrive at an accurate diagnosis in the overall management of patients with ongoing respiratory symptoms, but whether or not to treat with inhaled steroids may be better served by surrogate assessment of airway inflammation using FENO measurements.
Supported by the Lottery Grants Board of New Zealand and the Otago Respiratory Research Trust. GlaxoSmithKline (GSK) provided a personal educational grant to A.D.S. as a GSK Research Fellow, and also provided the study inhalers. This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org Conflict of Interest Statement: A.D.S. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. J.O.C. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. K.P.B. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. S.F. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. C.M. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. G.M.-S. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. G.P.H. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. D.R.T. has a potential conflict of interest in that Aerocrine, the manufacturer of NiOX analyzers, has recently agreed to donate NO analyzers to his laboratory for evaluation in the course of a future research study. This equipment was not available during the course of the present study, and Aerocrine has had no involvement in the design, execution, or analysis of the present study. Received in original form November 10, 2004; accepted in final form May 14, 2005
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
p < 0.01; 
p < 0.001).
